You are on page 1of 7

Outliers

Objective of
OLS

Leverage
points

Noninfluential
points

Influential
points

INFLUENTIAL POINTS
OUTLIERS
LEVERAGE POINTS WHICH CAN
INFLUENCE THE MODEL

Measures of influence

Cooks D-statistics

The Cooks distance statistics is a measure of distance between the least-squares


estimate based on all n observations in the model and the estimate obtained by
deleting the ith point.

DFFITS and DFBETAS


Indicates that how much the regression coefficient changes if the ith observation
were deleted. Such change is measured in terms of standard deviation units

Effect of influence points


Reduction in performance of the model.
Assumptions are usually not satisfied, hence influential points
should be removed first before jumping to the corrective
measure of the assumption validation.

Note: all influential points


should not be removed

R snapshot for checking the


influential points

You might also like